BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases12

SciCode (AA)

Artificial Analysis' own SciCode run. Shown for reference.

As of 24 Oct 2026, Claude Opus 5.5 leads SciCode (AA) on BenchLeader with 66.9%, ahead of Claude Opus 5.5 at 65.0%, across 232 model configurations with a published result.

Published by
Artificial Analysis
Category
Coding
Index weight
Reference only
Models
232
Data as of
24 Oct 2026
Newest result
23 Oct 2026

Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.

What the test looks like

SciCode gives models research-grade programming problems from physics, chemistry, biology and materials science, each broken into sub-steps that build a working scientific simulation or analysis.

How it is scored

Percent of main problems fully solved.

What to keep in mind

Artificial Analysis' own run of SciCode. Shown for reference; Epoch AI's run is counted in the index. See also the SciCode board.

Top 15
  1. 1Claude Opus 5.5 (max)66.9%
  2. 2Claude Opus 5.5 (xhigh)65.0%
  3. 3Claude Fable 5.1 (max)63.1%
  4. 4Muse Wonder Full (max)62.6%
  5. 5Gemini 4 Argon (high)61.8%
  6. 6Claude Fable 5 (max)61.0%
  7. 7Claude Sonnet 5.5 (max)61.0%
  8. 8Claude Fable 5.1 (xhigh)60.9%
  9. 9MiMo-V2.6-Pro60.9%
  10. 10Kimi K3.1 Pro60.8%
  11. 11Claude Opus 5.5 (high)60.4%
  12. 12Muse Wonder (max)60.3%
  13. 13Gemini 3.7 Flash (medium)59.8%
  14. 14Muse Spark 1.3 (xhigh)59.7%
  15. 15Kimi K3 (max)59.5%
232 of 232
#
1Claude Opus 5.5max66.9%72.62026-09-22
2Claude Opus 5.5xhigh65.0%70.02026-09-22
3Claude Fable 5.1max63.1%70.52026-09-01
4Muse Wonder Fullmax62.6%70.72026-10-12
5Gemini 4 Argonhigh61.8%70.82026-09-30
6Claude Fable 5max61.0%69.32026-06-09
7Claude Sonnet 5.5max61.0%69.32026-09-28
8Claude Fable 5.1xhigh60.9%69.52026-09-01
9MiMo-V2.6-Pro60.9%64.02026-09-21
10NewKimi K3.1 Pro60.8%66.82026-10-23
11Claude Opus 5.5high60.4%69.42026-09-22
12Muse Wondermax60.3%67.52026-10-12
13Gemini 3.7 Flashmedium59.8%63.12026-08-13
14Muse Spark 1.3xhigh59.7%64.92026-09-02
15Kimi K3max59.5%64.72026-07-16
16Claude Opus 5.5medium59.3%66.52026-09-22
17GLM 5.3max59.0%63.52026-08-18
18Step 5 Preview58.9%60.42026-09-18
19Muse Spark 1.3max58.8%67.12026-09-02
20Muse Spark 1.1xhigh58.8%61.62026-07-09
21Claude Fable 5.1high58.7%69.42026-09-01
22Gemini 3.1 Pro58.7%60.62026-02-19
23Claude Opus 5.5low58.6%64.72026-09-22
24Muse Spark Openxhigh58.5%62.02026-10-15
25GPT-5.6 Solhigh57.8%66.32026-07-09
26Grok 4.7high57.8%63.92026-09-21
27GPT-6 Solmax57.6%66.42026-09-22
28GPT-5.6 Solmedium57.4%64.42026-07-09
29Muse Spark 1.2xhigh57.4%63.22026-08-05
30Grok 4.7xhigh57.4%61.62026-09-21
31Claude Sonnet 5.5xhigh57.3%67.92026-09-28
32Gemini 3.7 Flashhigh57.2%63.52026-08-13
33GPT-5.6 Solmax57.1%68.02026-07-09
34GPT-5.6 Solxhigh57.1%66.62026-07-09
35Ling 3.1 Flash56.8%–2026-10-22
36Claude Fable 5.1low56.7%65.02026-09-01
37Gemini 3.8 Flashhigh56.6%64.42026-09-02
38GPT-6 Astramax56.5%70.42026-09-03
39Grok 4.6high56.5%62.92026-08-12
40Claude Opus 5max56.4%68.82026-07-24
41Claude Fable 5.1medium56.4%67.32026-09-01
42GPT-5.6 Sollow56.4%62.22026-07-09
43GPT-5.5high56.1%65.32026-04-23
44Grok 4.6medium55.9%64.22026-08-12
45NewDeepSeek V4.1 Flash 1017max55.9%64.02026-10-17
46GPT-6.1 Solhigh55.8%67.72026-09-29
47GPT-5.5xhigh55.8%66.12026-04-23
48GPT-6 Astraxhigh55.7%68.92026-09-03
49GPT-6.1 Solxhigh55.7%67.72026-09-29
50Claude Opus 5xhigh55.7%67.42026-07-24
51Gemini 3.7 Flashlow55.7%60.32026-08-13
52GPT-6 Astrahigh55.4%69.62026-09-03
53Claude Opus 5high55.4%67.92026-07-24
54Gemini 3.8 Flashmedium55.1%64.42026-09-02
55GPT-6 Solxhigh55.1%64.32026-09-22
56GPT-5.6 Terramax55.0%64.42026-07-09
57Gemini 3.8 Flashlow55.0%61.02026-09-02
58Claude Haiku 5.5max55.0%60.52026-10-07
59Grok 4.5high55.0%59.32026-07-08
60GPT-6 Solhigh54.9%63.72026-09-22

More coding benchmarks · All LLM benchmarks

Cite as: BenchLeader, “SciCode (AA) leaderboard”, data as of 24 Oct 2026, https://www.benchleader.com/benchmarks/aa_scicode.

SciCode (AA): questions

What does SciCode (AA) measure?
SciCode gives models research-grade programming problems from physics, chemistry, biology and materials science, each broken into sub-steps that build a working scientific simulation or analysis. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads SciCode (AA)?
Claude Opus 5.5 leads SciCode (AA) with 66.9% as of 24 Oct 2026, ahead of Claude Opus 5.5 at 65.0%.
How many models have SciCode (AA) results?
232 model configurations have a SciCode (AA) result on BenchLeader, all taken from Artificial Analysis.
Who runs SciCode (AA) and how often is it updated?
SciCode (AA) is published by Artificial Analysis. BenchLeader re-reads the published results once a day and records the date each result was published.
Does SciCode (AA) count toward the BenchLeader Index?
No. SciCode (AA) is shown for reference but left out of the composite index.